Journal of the American College of Cardiology Vol. 44, No. 4, 2004 © 2004 by the American College of Cardiology Foundation ISSN 0735-1097/04/$30.00 Published by Elsevier Inc. doi:10.1016/j.jacc.2004.06.042 EXPEDITED REVIEW Relationship Between Stress-Induced Myocardial and Measured by Coronary Calcium Tomography Daniel S. Berman, MD, FACC,*† Nathan D. Wong, PHD, FACC,‡ Heidi Gransar, MS,*† Romalisa Miranda-Peats, MPH,*† John Dahlbeck, BS,*† Sean W. Hayes, MD,*† John D. Friedman, MD, FACC,*† Xingping Kang, MD,*† Donna Polk, MD, MPH,*† Rory Hachamovitch, MD, FACC,§ Leslee Shaw, PHD,*† Alan Rozanski, MD, FACC*† Los Angeles and Irvine, California

OBJECTIVES We assessed the relationship between stress-induced myocardial ischemia on myocardial perfusion single-photon emission computed tomography (MPS) and magnitude of coronary artery calcification (CAC) by X-ray tomography in patients undergoing both tests. BACKGROUND There has been little evaluation regarding the relationship between CAC and inducible ischemia or parameters that might modify this relationship. METHODS A total of 1,195 patients without known coronary disease, 51% asymptomatic, underwent stress MPS and CAC tomography within 7.2 Ϯ 44.8 days. The frequency of ischemia by MPS was compared to the magnitude of CAC abnormality. RESULTS Among 76 patients with ischemic MPS, the CAC scores were Ͼ0 in 95%, Ն100 in 88%, and Ն400 in 68%. Of 1,119 normal MPS patients, CAC scores were Ͼ0, Ն100, and Ն400 in 78%, 56%, and 31%, respectively. The frequency of ischemic MPS was Ͻ2% with CAC scores Ͻ100 and increased progressively with CAC Ն100 (p for trend Ͻ0.0001). Patients with symptoms with CAC scores Ն400 had increased likelihood of MPS ischemia versus those without symptoms (p ϭ 0.025). Absolute rather than percentile CAC score was the most potent predictor of MPS ischemia by multivariable analysis. Importantly, 56% of patients with normal MPS had CAC scores Ն100. CONCLUSIONS Ischemic MPS is associated with a high likelihood of subclinical atherosclerosis by CAC, but is rarely seen for CAC scores Ͻ100. In most patients, low CAC scores appear to obviate the need for subsequent noninvasive testing. Normal MPS patients, however, frequently have extensive atherosclerosis by CAC criteria. These findings imply a potential role for applying CAC screening after MPS among patients manifesting normal MPS. (J Am Coll Cardiol 2004;44:923–30) © 2004 by the American College of Cardiology Foundation

An increasing body of literature demonstrates that measure- detecting CAD in patients with an intermediate likelihood ment of coronary artery calcification (CAC) by X-ray of this condition, and for nearly as long it has been established computed tomography (CT), using either electron beam as highly effective for risk stratification of patients with an computed tomography (EBCT) or multislice spiral com- intermediate or high likelihood of CAD (11–19). Thus, stress puted tomograpy (MSCT), represents a potent means for MPS is now commonly used for shaping key clinical manage- improving the diagnostic assessment and risk stratification ment decisions among patients with suspected or known of patients with suspected (CAD) CAD, such as distinguishing which CAD patients are likely to (1–10). Hence, the applications of this newer technology benefit from coronary revascularization versus medical man- may overlap some of the clinical applications associated with agement (20). The wide use and ubiquitous presence of noninvasive stress tests. For instance, for nearly three noninvasive stress tests, such as MPS, coupled with the decades, stress myocardial perfusion single-photon emission increasingly recognized utility and growing availability of CAC computed tomography (MPS) has been widely utilized for scanning, raises a new clinical problem for clinicians: how should CAC scanning be integrated with conventional stress imaging tests into the clinical assessment of patients with From the *Departments of Imaging and Medicine and the Burns and Allen Research Institute, Cedars-Sinai Medical Center, Los Angeles, California, †Depart- suspected and known CAD? Understanding the potential ment of Medicine, David Geffen School of Medicine, University of California, Los predictive relationship between CAC levels and the likelihood Angeles, California; ‡Heart Disease Prevention Program, University of California, of stress-induced myocardial ischemia would be central to Irvine, California; and the §Division of Cardiology, Keck School of Medicine, University of Southern California, Los Angeles, California. This study was supported addressing this question. Thus, we undertook the present study by a grant from the Eisner Foundation, Los Angeles, California. Dr. John J. to examine the potential inter-relationship between the pres- Mahamarian acted as Guest Editor for this paper. Manuscript received January 15, 2004; revised manuscript received June 14, 2004, ence and magnitude of CAC and the presence and magnitude accepted June 15, 2004. of inducible myocardial ischemia during stress MPS. 924 Berman et al. JACC Vol. 44, No. 4, 2004 MPS and EBCT August 18, 2004:923–30

reduced workload. Tc-99m sestamibi MPS imaging was Abbreviations and Acronyms begun 15 to 30 min after radioisotope injection (21). CAC ϭ coronary artery calcium Adenosine MPS protocol. In 146 (12%) of the study CAD ϭ coronary artery disease subjects, adenosine stress was performed (22). Patients were ϭ CT computed tomography instructed not to consume caffeine products for 24 h before EBCT ϭ electron beam computed tomography ϭ MPS. Following rest MPS, adenosine was infused (140 HU Hounsfield units ␮ MPS ϭ myocardial perfusion single-photon emission g/kg/min for 5 to 6 min), and Tc-99m sestamibi was computed tomography injected at the end of the 2nd or 3rd min of infusion for the MSCT ϭ multislice spiral computed tomography 5- and 6-min infusions, respectively. In patients who could ϭ SDS summed difference score tolerate it, low-level treadmill exercise, as an adjunct to SPECT ϭ single-photon emission computed tomography SRS ϭ summed rest score adenosine infusion, was performed at 0% to 10% grade, at 1 SSS ϭ summed stress score to 1.7 miles/h. The Tc-99m sestamibi MPS was initiated Tc ϭ technetium approximately 60 min after the end of adenosine infusion in Tl ϭ thallium patients who did not exercise and 15 to 60 min after injection in those with adjunctive exercise. METHODS During both types of stress, blood pressure was recorded at rest, at the end of each stress stage, and at peak stress. We evaluated 1,195 patients who underwent rest/stress dual Maximal ST-segment change was assessed as horizontal, isotope MPS and CAC scanning by EBCT (Imatron upsloping, or downsloping, and electrocardiographic ischemia C-150 or GE e-Speed, GE-Imatron Inc., South San was defined as ST-segments Ն1 mm horizontal or downslop- Francisco, California) or MSCT (Siemens Volume Zoom, ing or Ն1.5 mm upsloping at 80 ms after the J point. Siemens Medical Systems, Forchheim, Germany) at SPECT acquisition protocol. The MPS studies were Cedars-Sinai Medical Center within six months of each performed on multidetector scintillation cameras using an other (7.2 Ϯ 44.8 days). The mean age of the study elliptical 180° acquisition for 60 to 64 projections at 20 s per population was 58.4 Ϯ 10.3 years, and 869 (72.7 %) of the projection (21). For Tl-201, two energy windows were used, patients were male. Patients underwent MPS on a clinical including a 30% window centered on the 68- to 80-keV basis, and CAC imaging was performed either on a basis of peak and a 10% window centered on the 167 keV peak. For self-referral (n ϭ 94 patients), physician-referral (n ϭ 777 Tc-99m sestamibi, a 15% window centered on the 140-keV patients), or ongoing research (n ϭ 324 patients) in the peak was used, and images were obtained in both supine and Early Identification of Subclinical Atherosclerosis by Non- prone positions. For supine rest and stress MPS studies, gated invasive Imaging Research (EISNER) study. Exclusion SPECT was performed, obtaining 8 to 16 frames/cycle. criteria included prior coronary bypass surgery or percuta- Images were acquired using a 64 ϫ 64 image matrix and were neous coronary intervention, history of myocardial infarc- subject to quality control measures as previously described (21). tion, known valvular heart disease, or primary cardiomyop- No attenuation or scatter correction was employed. athy. This research was approved by the Cedars-Sinai Interpretation of SPECT. Semiquantitative visual interpre- Medical Center Institutional Review Board. tation was performed using 20 segments for each image set. Imaging and stress protocol. Patients were injected intra- Segments were scored by consensus of two experienced ob- venously at rest with thallium-201 (Tl-201) (3.0 to 4.5 servers using a 5-point score (0 ϭ normal, 1 ϭ equivocal, 2 ϭ mCi) with dose variation based on patient weight. Rest moderate, 3 ϭ severe reduction of radioisotope uptake, and 4 Tl-201 SPECT was initiated 10 min after injection of the ϭ absence of detectable tracer uptake in a segment) (21). radionuclide (21). Scintigraphic indices. The summed stress score (SSS) and Exercise MPS protocol. Following rest MPS, symptom- summed rest score (SRS) were obtained by adding the limited Bruce protocol treadmill exercise testing was per- scores of the 20 segments of the respective images (19). An formed in 1,049 (88%) of the 1,195 study subjects. Exercise SSS Ն4 was considered abnormal (21). The sum of the end-points included physical exhaustion, severe , differences between each of the 20 segments from these sustained ventricular tachycardia, hemodynamically signifi- images was defined as the summed difference score (SDS), cant supraventricular dysrhythmias, or exertional hypoten- incorporating assessment of both the extent and severity of sion. In accordance with our policy to discontinue anti- stress-induced myocardial ischemia, each of which indepen- ischemic medications before exercise testing, only 36 dently adds prognostic information (11). The SDS was subjects (3.0%) were on beta-blocking medication Ͻ48 h, converted to percent myocardium ischemic by dividing the 10 (0.8%) were on calcium blockers Ͻ24 h, and 3 (0.3%) SDSby80—maximum potential score (4 ϫ 20)—and were on nitrates Ͻ6 h before MPS. At near-maximal multiplying by 100. Ischemic and moderate-to-severe isch- exercise, technetium-99m (Tc-99m) sestamibi (25 to 40 emic MPS studies were defined by Ն5% and Ͼ10% of the mCi) was injected (actual patient dose varied with patient myocardium, respectively (20). Because the presence and weight) and exercise continued for an additional 1 min after magnitude of hypoperfusion induced during adenosine injection at peak workload, and an additional 1 to 2 min at MPS is roughly equivalent to the magnitude of ischemia JACC Vol. 44, No. 4, 2004 Berman et al. 925 August 18, 2004:923–30 MPS and EBCT induced during maximal exercise (23–27), we employ the relationships between CAC score category and likelihood of term “inducible ischemia” to represent both the ischemia an ischemic MPS differed by these factors. Analyses were during exercise and adenosine MPS, even though stress conducted similarly across CAC score age- and gender- actual ischemia is often not produced by the adenosine. adjusted percentile levels (0 to 24th, 25th to 49th, 50th to Calcium scanning. The imaging protocol involved acquir- 74th, 75th to 89th, and 90th to 99th). ing a single scan on each patient, consisting of approxi- Moreover, among each of these percentile categories, the mately 30 to 40 3- or 2.5-mm slices for EBCT and MSCT relation of CAC score category with MPS result was also respectively, sufficient to cover the entire heart, with trig- assessed. The non-parametric Wilcoxon rank sum test for gering at 50% to 80% of the cardiac cycle. Breath-holding proportional variables and the Pearson chi-square test of instructions were given to minimize misregistration. association for categorical variables were used to compare CALCIUM SCAN INTERPRETATION. Foci of CAC were iden- subjects with an abnormal versus normal SPECT result. In tified by an experienced radiographic technologist and addition, the Fisher exact test was employed when appro- scored using semiautomatic commercial software on a priate. Multiple logistic regression was performed to exam- NetraMD workstation (ScImage, Los Altos, California) by ine whether log (CAC score ϩ1) was independently asso- detection of at least three contiguous pixels (voxel size ϭ ciated with the likelihood of an abnormal MPS result, after 1.03 mm3) of peak density Ն130 Hounsfield units (HU) adjustment for standardized age, gender, presence of symp- within a coronary artery. Scoring was performed and verified toms, other clinical characteristics, and CAD risk factors. by an experienced imaging cardiologist. The software cal- The number of days between tests was included to control culated lesion-specific scores as the product of the area of for the progression of CAD that might occur during the up each calcified focus and peak CT number (scored as 1 if 131 to six-month interval between tests. Receiver operating to 199 HU, 2 if 200 to 299 HU, 3 if 300 to 399 HU, and characteristic (ROC) analysis was used to compare the 4 if 400 HU or greater) according to the Agatston method information in predicting ischemic MPS based on age and (28). These were summed across all lesions identified within gender; age, gender, and symptoms; and the combination of left main, left anterior descending, left circumflex, and right these with the CAC score. All continuous variables are to provide arterial-specific calcium scores, expressed as mean Ϯ SD. and across arteries to provide the total CAC score that was used as the principal EBCT/CT measurements in this study. RESULTS The CAC percentile score based on age and gender was assigned based on cut points from a large database of Raggi et Pertinent clinical characteristics for our study population, al. (4) programmed into the ScImage calcium scoring software. stratified by MPS result are shown in Table 1.Of79 Historical and other clinical variables. Each patient was patients with an abnormal MPS, 76 (96%) also had Ն5% questioned regarding the following: chest symptoms, di- ischemia. The remaining 3 with abnormal MPS had SSS ϭ vided into four categories (asymptomatic, non- 4 but Ͻ5% ischemia and were combined with the other anginal chest pain, atypical, and typical angina) (29); the 1,116 patients with normal MPS to constitute our normal presence or absence of ; medication use; and MPS group. Compared to the patients with a normal MPS, the following historical coronary risk factors: 1) family history patients with an ischemic study had a more abnormal of early coronary heart disease (male primary relative Ͻ55 years coronary risk profile (being significantly older, with more of age, female Ͻ65 years of age); 2) diabetes; 3) hypertension; males, more hypertension, and a higher mean pre-test 4) high serum cholesterol level; and 5) current or past usage of likelihood of CAD) and significantly greater functional test cigarettes. Weight and height were measured with body mass abnormalities, including more exercise-induced chest pain, index calculated as weight (kg)/height m2. Bayesian analyses of shorter exercise duration, more exercise-induced ST- patient age, gender, symptoms, risk factors, and, in the exercise segment depression, and a lower mean peak exercise heart cohort, exercise results were used to calcu- rate. In addition, the ischemic MPS patient group mani- late the pre-MPS likelihood of CAD according to a computer fested a significantly higher mean CAC score value versus program (CADENZA) (30). the normal MPS patient group. Statistical analyses. Total calcium scores were classified Comparison of MPS and CAC scanning. Figure 1 illus- into six categories: 0 (calcium absent); 1 to 9 (minimal); 10 trates both the frequency of an ischemic MPS and a to 99; 100 to 399; 400 to 999; and Ն1,000 (5). When being moderate to severe ischemic MPS, according to the absolute analyzed as a continuous variable, a log transform of CAC CAC score, grouped into six categories. Below a calcium was used to reduce the amount of skew of CAC scores and score of 100, the frequency of an ischemic MPS was very allow for multivariable statistical analyses that are based on low (Ͻ2% overall). At the other end of the spectrum, among the assumption of normality. The prevalences of an isch- patients with calcium scores Ͼ1,000, one-fifth had an emic and moderate to severe ischemic MPS were compared ischemic MPS, with less than one-half of these studies across calcium score categories using the chi-square test of (8.6% of all patients with CAC score Ͼ1,000) demonstrat- trend. These analyses were also run with stratification by ing moderate to severe MPS ischemia. gender and by presence of symptoms to examine whether Comparison of the SDS by MPS versus log-transformed 926 Berman et al. JACC Vol. 44, No. 4, 2004 MPS and EBCT August 18, 2004:923–30

Table 1. Characteristics of the Study Population Overall Normal MPS Ischemic MPS p Values (76 ؍ n) *(1,119 ؍ n) (1,195 ؍ Parameter (n Age, yrs 58.4 Ϯ 10.3 58.1 Ϯ 10.2 61.9 Ϯ 11.1 0.0028 Female 326 (27.3%) 318 (28.4%) 8 (10.5%) 0.001 Symptom class Asymtomatic 609 (51.0%) 572 (51.1%) 37 (48.7%) 0.681 Asymptomatic/SOB 658 (55.1%) 621 (55.5%) 37 (48.7%) 0.248 Non-anginal pain 112 (9.4%) 108 (9.7%) 4 (5.3%) 0.204 Atypical angina 380 (31.8%) 359 (32.1%) 21 (27.6%) 0.420 Typical angina 45 (3.8%) 31 (2.8%) 14 (18.4%) Ͻ0.0001 Risk factors High cholesterol 897 (75.1%) 844 (75.4%) 53 (69.7%) 0.267 Hypertension 514 (43.0%) 465 (41.6%) 49 (64.5%) Ͻ0.0001 Family history 448 (37.5%) 414 (37.0%) 34 (44.7%) 0.177 Diabetes 138 (11.6%) 126 (11.3%) 12 (15.8%) 0.232 Smoking 84 (7.0%) 78 (7.0%) 6 (7.9%) 0.760 BMI 26.9 Ϯ 4.8 26.8 Ϯ 4.7 27.9 Ϯ 5.2 0.1028 CAC score 438.9 Ϯ 703.4 388.9 Ϯ 617.0 1,175.0 Ϯ 1,272.0 Ͻ0.0001 Pre-rest likelihood of CAD (%) 24.4 Ϯ 24.1 23.6 Ϯ 23.1 36.6 Ϯ 33.6 0.0019 Exercise stress test* n ϭ 1,049 n ϭ 989 n ϭ 58 Ischemic stress ECG 220 (21.0%) 184 (18.6%) 36 (62.1%) Ͻ0.0001 Exercise time 9.3 Ϯ 2.7 9.4 Ϯ 2.7 7.9 Ϯ 2.8 0.0001 Exercise chest pain 83 (7.9%) 62 (6.3%) 21 (36.2%) Ͻ0.0001 Rest HR 68.2 Ϯ 12.2 68.3 Ϯ 12.1 67.7 Ϯ 13.9 0.738 Peak HR 154.3 Ϯ 14.8 155 Ϯ 14.4 142.1 Ϯ 16.6 Ͻ0.0001 Rest SBP 136.7 Ϯ 18.6 136.4 Ϯ 18.7 142.1 Ϯ 17.7 0.0209 Peak SBP 171.6 Ϯ 20.7 171.8 Ϯ 20.8 167.8 Ϯ 18.4 0.0910 Rest DBP 80.4 Ϯ 9.3 80.3 Ϯ 9.2 83.3 Ϯ 9.6 0.0250 Peak DBP 77.2 Ϯ 9.7 77.0 Ϯ 9.6 81.0 Ϯ 10.9 0.0017

*Adenosine stress excluded. BMI ϭ body mass index; CAC ϭ coronary artery calcium; CAD ϭ coronary artery disease; HR ϭ heart rate; MPS ϭ myocardial perfusion single-photon emission computed tomography; SBP ϭ systolic blood pressure; SOB ϭ shortness of breath.

CAC scores revealed a significant but only fair Spearman CAC percentile groupings (p Ͻ 0.0005). For patients below correlation (r ϭ 0.195, p Ͻ 0.05). This relationship was the 50th percentile of CAC abnormality, the frequency of an largely governed by a wide distribution of CAC scores ischemic MPS study was very low (Ͻ2%). Once patients among patients with normal MPS, as illustrated in Figure 2. reached the 50th percentile, the frequency of ischemic MPS Of patients with ischemic MPS (Ն5% ischemia), 88% had was substantially increased, but it did not vary substantially CAC scores Ն 100. Importantly, 56% of the patients with among 50th to 74th, 75th to 89th, and Ն90th percentile Ն normal MPS also had CAC scores 100. groups. Assessment according to CAC percentile score. Because Figure 4 demonstrates the frequency of an ischemic MPS both age and gender are known to impact CAC, we as a function of both the absolute CAC score and the CAC examined the relationship between the age-gender percen- score (for all values above a 50th percentile CAC ranking). tile CAC score and the frequency of an ischemic MPS (Fig. For each of the shown CAC percentile groups, there was a 3). As with the absolute CAC score, a significant overall trend existed for the frequency of MPS ischemia across

Figure 1. The frequency of an ischemic myocardial perfusion single- Figure 2. Distribution of coronary artery calcium (CAC) scores for the photon emission computed tomography (Ն5% ischemic) (gray bars) and of 1,119 patients manifesting a normal myocardial perfusion single-photon a moderate to severe ischemia (Ͼ10% ischemic) (black bars) for patients emission computed tomography (MPS) (left) and the 76 patients with an divided into six coronary artery calcium (CAC) score groupings. ischemic MPS (right). JACC Vol. 44, No. 4, 2004 Berman et al. 927 August 18, 2004:923–30 MPS and EBCT

Figure 3. The frequency of an ischemic myocardial perfusion single- Figure 5. The frequency of an ischemic myocardial perfusion single- photon emission computed tomography (MPS) (Ն5% ischemic) (gray photon emission computed tomography for each of the six coronary artery bars) and moderate to severe ischemic MPS (Ͼ10% ischemic) (black bars) calcium (CAC) score subgroups, further subdivided on the basis of according to five groupings of age-gender adjusted coronary artery calcium symptoms (chest pain and/or shortness of breath) being absent (gray bars) percentile score. or present (black bars). stepwise increase in the frequency of an ischemic SPECT the CT scan, respectively. Interaction terms with test order study as the absolute CAC level increased from low to high and CAC score or category in relation to likelihood of values. Patients with a high percentile ranking (owing to positive MPS were nonsignificant (p Ͼ 0.05), suggesting relatively younger age) but low absolute CAC score did not there was no association of test order and the relation manifest inducible ischemia during MPS. between CAC and likelihood of MPS ischemia. The impact of symptoms. Dividing patients on the basis Multivariate predictors of abnormal MPS result. Multi- of presence or absence of symptoms altered the observed variable logistic regression analysis revealed that the log frequencies of ischemic MPS studies among patients whose CAC score was the most potent predictor of MPS ischemia calcium scores exceeded 10, as demonstrated in Figure 5.As in our study (Table 2). Additional significant variables the absolute CAC score increased, the difference in the included gender, the presence of symptoms (i.e., chest pain frequency of MPS ischemia among symptomatic versus and/or shortness of breath), a history of hypertension and asymptomatic subjects became progressively more pro- high cholesterol, and increased body mass index. To assess nounced. When the CAC score was high (Ն400), a the potential incremental information provided by CAC significantly higher frequency of ischemia occurred in the scanning for predicting MPS ischemia, we performed an symptomatic than in the asymptomatic group (18.5% vs. ROC analysis in which the clinical variables of age and 10.4%; p ϭ 0.025). gender were first forced into the multivariate prediction of Sequence of testing. In 37.8%, 15.6%, and 46.5% of MPS ischemia (Fig. 6). The subsequent addition of coro- patients the nuclear scan came before, the same day, or after nary risk factors and chest pain symptom information then significantly increased the size of the ROC area over age and gender from 67% to 74% (incremental p ϭ 0.0061). Finally, the further addition of the CAC score significantly in- creased the resultant ROC area beyond the previous curve to 80% (incremental p ϭ 0.0051). All three ROC curves were highly significant (p Ͻ 0.0001). Thus, the CAC score adds incremental information for predicting the likelihood

Table 2. Multivariate Predictors of a Ն5% Ischemic Myocardial Perfusion Single-Photon Emission Computed Tomography Odds Predictor* Ratio p Value 95% CI Log CAC 3.36 0.000 2.13–5.29 Figure 4. The frequency of an abnormal myocardial perfusion single- Age 1.18 0.260 0.88–1.59 photon emission computed tomography (MPS) study according to coro- Male gender 1.54 0.020 1.07–2.22 nary artery calcium (CAC) percent ranking and absolute CAC score History of high BP 1.41 0.010 1.09–1.82 (patients with percentile rankings Ͻ50% are not shown owing to the very History of high cholesterol 0.72 0.009 0.57–0.92 low frequency of abnormal MPS studies in such patients). Within each of Family history of CAD 1.34 0.026 1.04–1.73 three CAC percentile groups, patients are further divided on the basis of History of diabetes 0.98 0.829 0.78–1.22 their absolute CAC score, condensed into three subgroups: CAC scores of History of smoking 1.06 0.631 0.84–1.35 Ն 1 to 99, 100 to 399, and 400. The first set of numbers below each bar BMI 1.30 0.049 1.00–1.70 represents the absolute CAC score; the second set represents the mean age Symptoms 1.40 0.011 1.08–1.81 Ϯ SD; and the third set represents the number of patients within each of the nine subgroups that are illustrated. Note that regardless of percentile Days 0.92 0.552 0.70–1.21 ranking, the frequency of an ischemic MPS study was relatively high when BP ϭ blood pressure; BMI ϭ body mass index; CAC ϭ coronary artery calcium; CI the absolute CAC score was Ն400, and relatively low when the CAC score ϭ confidence interval; CAD ϭ coronary artery disease; Days ϭ number of days was Ͻ100. between tests. *Predictors have been standardized. 928 Berman et al. JACC Vol. 44, No. 4, 2004 MPS and EBCT August 18, 2004:923–30

ment of atherosclerotic burden by CAC testing may be useful in assessing of the need for aggressive attempts to prevent coronary events. Prior studies. Despite the apparent overlap by noninvasive stress tests (11–20) and CAC scanning (1–10) for assessing outcomes, few investigations have inquired into the inter- relationship between results of stress imaging and the presence and magnitude of CAC abnormality. With one exception, the small number of prior investigations regard- ing myocardial ischemia and CAC abnormality primarily focused on how these indices compared in their ability to predict angiographic coronary stenoses, without direct com- parison of CAC to myocardial ischemia per se (32–35). Figure 6. Receiver operating characteristic curve analysis showing incre- Only one prior study has specifically focused on the inter- mental prognostic value of coronary artery calcium score in predicting likelihood of Ն5% ischemia. CCS ϭ coronary calcium score; CRF ϭ relationship between the CAC score and myocardial hypo- coronary risk factors; Sx ϭ symptoms. perfusion by MPS (36). In that report, He et al. (36) noted a similar threshold phenomenon with almost no observable of MPS ischemia, even after other predictors of MPS myocardial hypoperfusion among patients with a CAC ischemia are first considered. score Ͻ100 and with a marked increase in the frequency of DISCUSSION an abnormal MPS in patients with high CAC values. Contrasting with the study of He et al. (36), our study Our results indicate that a threshold phenomenon governs showed a much lower frequency of MPS abnormality the relationship between the extent of calcified plaque as among patients with CAC scores Ն400. The CAC mea- measured by the CAC score by X-ray computed tomogra- surements, per se, are highly standardized, and the distri- phy and the presence of myocardial ischemia, as measured bution of symptoms and risk factors were not very dissimilar by stress MPS. When grouped according to calcium score, for these two studies. Accordingly, the observed differences both patients with a relatively low and relatively high in the frequency of MPS abnormality among patients with likelihood of demonstrating an ischemia MPS study could high CAC scores were most likely due to either differences be identified. Specifically, among the patients with a cal- in the interpretative criteria used to assess the MPS results cium score Ͻ100 in our study, MPS ischemia was rare, in the two studies and/or differences in the referral pattern occurring in Ͻ2% of such patients. This low frequency of to MPS testing after CAC scanning. A preliminary study ischemia with a CAC score Ͻ100 was present in patients assessing the frequency of MPS abnormality among 121 with and without clinical symptoms, although a trend patients with high CAC scores reported a frequency of toward more ischemia in symptomatic patients with scores MPS abnormality that parallels that noted in our study (37). 10 to 99 was observed. As the CAC score increased in The impact of clinical symptoms. Most clinical studies magnitude above 100, the frequency of myocardial ischemia on regarding coronary artery CT have looked primarily at MPS increased progressively. Among patients with CAC asymptomatic subjects. In our study, however, nearly one- scores exceeding 1,000, 20% manifested ischemia by MPS. half of our patients had clinical symptoms, ranging from Our results further indicate that the likelihood of myo- patients complaining of shortness of breath to patients cardial ischemia by MPS is more tightly related to the presenting with typical angina. Like the asymptomatic absolute CAC score rather than age-gender–stratified CAC patients, our symptomatic patients, as a group, had a very percentile score. For example, among patients with CAC low frequency of MPS ischemia when the CAC score was score exceeding 400, the frequency of myocardial ischemia Ͻ10. But for each higher CAC score subgroup, including was comparably high over a wide range of percentile even patients with a CAC score of 10 to 99, the symptomatic rankings. These data indicate an important distinction. patients manifested an approximate tripling of the frequency of Whereas a low CAC score with a high percentile ranking in MPS ischemia compared to patients without clinical symp- young patients may be indicative of long-term risk for toms. Accordingly, these results suggest that the presence of developing cardiac events (4,6,31), this same score is prob- clinical symptoms is a strong modifier of the relationship ably not predictive of short-term risk, given the finding that between CAC scores and the likelihood of inducible ischemia most such patients have no evidence of ischemia on MPS. once any substantial degree of CAC is present. Thus, further testing by MPS of patients found to have high Multivariable analysis. The presence of coronary calcium CAC percentile but a CAC score Ͻ100 would not appear to was the most potent multivariable predictor of myocardial be needed in most patients. Of additional interest, our ischemia during MPS. In addition, chest pain, gender, and findings demonstrated that a relatively high proportion of certain coronary risk factors were additional significant patients referred for MPS and found to have no MPS multivariate predictors of myocardial ischemia. Neverthe- ischemia have a CAC score Ն100, suggesting that assess- less, even when all other predictors of SPECT abnormality JACC Vol. 44, No. 4, 2004 Berman et al. 929 August 18, 2004:923–30 MPS and EBCT were forced into a multivariate model for predicting MPS that within this range of CAC scores, clinical factors such as ischemia, the CAC score still added significant incremental symptoms, gender, coronary risk factors, and stress ECG value for this prediction. These results thus parallel prior results may serve to impact substantially on the observed observations indicating that CAC imaging is a potent frequency of MPS ischemia. Accordingly, future studies incremental predictor for cardiac events, over and above the involving large number of patients with CAC scores in the information provided by other known predictors of coronary range of 100 to 400 are needed to define the best combi- events and all-cause mortality (5,6,10). nation of clinical predictors for predicting an ischemic MPS Study limitations. Most studies involving tomographic study in this CAC score range. In general, however, it measurements of coronary calcium have been limited pri- appears from our data that within this range, any threshold marily to asymptomatic subjects or patients. Various factors for referral to MPS testing following CAC scanning should have accounted for this, including the relative lack of be lower among symptomatic patients or among those of widespread third-party reimbursement for this procedure, male gender. Conversely, an asymptomatic presentation is leading to out-of-pocket payment by largely asymptomatic likely to be associated with a higher recommended CAC patients, and the established practice of direct stress testing threshold value before referral for stress testing becomes referral among patients with anginal symptoms. Thus, even cost-effective after CAC scanning. though one-half our population was symptomatic, the One potential approach to determining whether MPS population was heavily skewed toward atypical chest pain would be of value in a given patient might be to combine the symptoms and a relatively low likelihood of CAD, averaging CAC score with all other relevant clinical and historical Ͻ 25% CAD likelihood. information into a Bayesian estimate of likelihood for Additionally, one-half of the patients were asymptomatic. angiographically significant CAD, as suggested by a recent Although all patients in this study were referred for MPS by study (39). However, more work is needed to validate this their physicians, it should be noted that there are currently approach and compare it with other potential approaches no class Ia indications for stress MPS in asymptomatic for integrating the results of CAC scanning into clinical individuals (38). It is possible that some of the inter- practice, such as the use of the Framingham risk score in relationships noted in this study, such as the relative asymptomatic patients. strength of chest pain symptoms to predict myocardial Third, the wide range of CAC scores in our patients with ischemia, would have been affected had we been able to normal MPS studies exposes an important limitation rele- incorporate more patients with typical angina into this vant to all forms of stress testing: they do not effectively study. Similarly, our ability to perform subanalyses, such as screen for subclinical atherosclerosis. For instance, only 22% the impact of symptoms and gender effects, had limited of our patients with a normal SPECT study had no evidence statistical power owing to the relatively low frequency of of CAC, 56% had a CAC score Ն100, and 31% had a CAC MPS ischemia in our study. Accordingly, prospective stud- score Ն400. Further, even among those with CAC scores ies that focus on the inter-relationship between MPS Ն1,000, 85% of asymptomatic and nearly 68% of symptom- studies and calcium scores in populations containing more atic patients had a normal MPS study. patients with typical angina and/or inducible myocardial Along these lines, there are yet no available data to ischemia would be useful for extending our findings. compare the relative short and long-term risk for cardiac Finally, as CAC matures as a testing modality in cardi- events among patients with various combinations of MPS ology, it will be interesting to evaluate whether the per- ceived accuracy of this test is subject to patient referral bias, results and CAC scores, such as those presenting with the as has been observed for other noninvasive tests in cardiol- combination of very high CAC scores but normal MPS ogy (39). results. It is reasonable to hypothesize that such patients Clinical implications. Three broad implications emanate might be at low short-term risk but high long-term risk for from our results. First, they help to define the future cardiac events. If so, CAC could then be unmasking a indications for stress MPS referral after CAC imaging. subgroup of patients who would receive more aggressive Specifically, it appears that the referral of patients for MPS anti-atherosclerotic intervention than would have been in- is generally not needed when the CAC score is Ͻ100 due to dicated based on the results of MPS testing alone. Accord- the very low likelihood of observing inducible myocardial ingly, future studies incorporating the prognostic follow-up ischemia in such patients. Conversely, when the CAC score data from patients undergoing both studies would now be of exceeds 400, stress imaging would appear to be generally interest, so as to determine which patients with normal beneficial, because the frequency of inducible ischemia is stress imaging tests are best suited for undergoing subsequent substantial within this CAC range, even in asymptomatic CAC scanning. patients. Second, our results indicate that CAC scores in the range Reprint requests and correspondence: Dr. Daniel S. Berman, of 100 to 400 constitute a relatively large “gray zone” relative Director of Cardiac Imaging, Cedars-Sinai Medical Center, 8700 to the issue of who may require stress-test referral following Beverly Building, Room 1258, Los Angeles, California 90048. CAC imaging. Pending confirmation, our results suggest E-mail: [email protected] 930 Berman et al. JACC Vol. 44, No. 4, 2004 MPS and EBCT August 18, 2004:923–30

REFERENCES artery disease undergoing stress myocardial perfusion single photon emission computed tomography. Circulation 2003;107:2900–7. 1. Arad Y, Spadaro LA, Goodman K, et al. Predictive value of electron beam 21. Berman DS, Kiat H, Friedman JD, et al. Separate acquisition rest computed tomography of the coronary arteries. 19-month follow-up of thallium-201/stress technetium-99m sestamibi dual-isotope myocar- 1,173 asymptomatic subjects. Circulation 1996;93:1951–3. dial perfusion single-photon emission computed tomography: a clini- 2. Arad Y, Spadaro LA, Goodman K, Newstein D, Guerci AD. cal validation study. J Am Coll Cardiol 1993;22:1455–64. Prediction of coronary events with electron beam computed tomogra- 22. Berman DS, Kang X, Hayes SW, et al. Adenosine myocardial phy. J Am Coll Cardiol 2000;36:1253–60. perfusion single-photon emission computed tomography in women 3. Wong ND, Hsu JC, Detrano RC, Diamond G, Eisenberg H, Gardin compared with men. Impact of diabetes mellitus on incremental JM. Coronary artery calcium evaluation by electron beam computed prognostic value and effect on patient management. J Am Coll Cardiol tomography and its relation to new cardiovascular events. Am J 2003;41:1125–33. Cardiol 2000;86:495–8. 23. Albro PC, Gould KL, Westcott RJ, Hamilton GW, Ritchie JL, 4. Raggi P, Callister TQ, Cooil B, et al. Identification of patients at Williams DL. Noninvasive assessment of coronary stenoses by myo- increased risk of first unheralded acute by cardial imaging during pharmacologic coronary vasodilatation. III. electron-beam computed tomography. Circulation 2000;101:850–5. Clinical trial. Am J Cardiol 1978;42:751–60. 5. Keelan PC, Bielak LF, Ashai K, et al. Long-term prognostic value of 24. Timmis AD, Lutkin JE, Fenney LJ, et al. Comparison of dipyrid- coronary calcification detected by electron-beam computed tomogra- amole and treadmill exercise for enhancing thallium-201 perfusion phy in patients undergoing coronary angiography. Circulation 2001; defects in patients with coronary artery disease. Eur Heart J 1980;1: 104:412–7. 275–80. 6. Raggi P, Cooil B, Callister TQ. Use of electron beam tomography data 25. Narita M, Kurihara T, Usami M. Noninvasive detection of coronary to develop models for prediction of hard coronary events. Am Heart J artery disease by myocardial imaging with thallium-201—the signifi- 2001;141:375–82. cance of pharmacologic interventions. Jpn Circ J 1981;45:127–40. 7. Park R, Detrano R, Xiang M, et al. Combined use of computed 26. Leppo JA. Dipyridamole-thallium imaging: the lazy man’s stress test. tomography coronary calcium scores and C-reactive protein levels in J Nucl Med 1989;30:281–7. predicting cardiovascular events in nondiabetic individuals. Circulation 27. Varma SK, Watson DD, Beller GA. Quantitative comparison of 2002;106:2073–7. thallium-201 scintigraphy after exercise and dipyridamole in coronary 8. Vliegenthart R, Oudkerk M, Song B, van der Kuip DA, Hofman A, artery disease. Am J Cardiol 1989;64:871–7. Witteman JC. Coronary calcification detected by electron-beam com- 28. Wong ND, Sciammarella MG, Polk D, et al. The metabolic syn- puted tomography and myocardial infarction. The Rotterdam Coro- drome, diabetes, and subclinical atherosclerosis assessed by coronary nary Calcification Study. Eur Heart J 2002;23:1596–603. calcium. J Am Coll Cardiol 2003;41:1547–53. 9. Kondos GT, Hoff JA, Sevrukov A, et al. Electron-beam tomography 29. Diamond GA, Forrester JS, Hirsch M, et al. Application of condi- coronary artery calcium and cardiac events: a 37-month follow-up of tional probability analysis to the clinical diagnosis of coronary artery 5,635 initially asymptomatic low- to intermediate-risk adults. Circu- disease. J Clin Invest 1980;65:1210–21. lation 2003;107:2571–6. 30. Diamond GA, Staniloff HM, Forrester JS, Pollock BH, Swan HJ. 10. Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ. Computer-assisted diagnosis in the noninvasive evaluation of patients Prognostic value of cardiac risk factors and coronary artery calcium with suspected coronary artery disease. J Am Coll Cardiol 1983;1:444–55. screening for all-cause mortality. Radiology 2003;228:826–33. 31. Raggi P. Prognostic implications of absolute and relative calcium 11. Ladenheim ML, Pollock BH, Rozanski A, et al. Extent and severity of scores. Herz 2001;26:252–9. myocardial hypoperfusion as predictors of prognosis in patients with 32. Kajinami K, Seki H, Takekoshi N, Mabuchi H. Noninvasive predic- suspected coronary artery disease. J Am Coll Cardiol 1986;7:464–71. tion of coronary atherosclerosis by quantification of coronary artery 12. Staniloff HM, Forrester JS, Berman DS, Swan HJ. Prediction of calcification using electron beam computed tomography: comparison death, myocardial infarction, and worsening chest pain using thallium with electrocardiographic and thallium exercise stress test results. J Am scintigraphy and exercise electrocardiography. J Nucl Med 1986;27: Coll Cardiol 1995;26:1209–21. 1842–8. 33. Schmermund A, Baumgart D, Sack S, et al. Assessment of coronary 13. Brown KA. Prognostic value of thallium-201 myocardial perfusion calcification by electron-beam computed tomography in symptomatic imaging. A diagnostic tool comes of age. Circulation 1991;83:363–81. patients with normal, abnormal or equivocal exercise stress test. Eur 14. Iskandrian AS, Chae SC, Heo J, Stanberry CD, Wasserleben V, Cave Heart J 2000;21:1674–82. V. Independent and incremental prognostic value of exercise single- photon emission computed tomographic (SPECT) thallium imaging 34. Shavelle DM, Budoff MJ, LaMont DH, Shavelle RM, Kennedy JM, in coronary artery disease. J Am Coll Cardiol 1993;22:665–70. Brundage BH. Exercise testing and electron beam computed tomog- 15. Heller GV, Herman SD, Travin MI, Baron JI, Santos-Ocampo C, raphy in the evaluation of coronary artery disease. J Am Coll Cardiol McClellan JR. Independent prognostic value of intravenous dipyrid- 2000;36:32–8. amole with technetium-99m sestamibi tomographic imaging in pre- 35. Schmermund A, Denktas AE, Rumberger JA, et al. Independent and dicting cardiac events and cardiac-related hospital admissions. J Am incremental value of coronary artery calcium for predicting the extent Coll Cardiol 1995;26:1202–8. of angiographic coronary artery disease: comparison with cardiac risk 16. Hachamovitch R, Berman DS, Kiat H, et al. Exercise myocardial factors and radionuclide perfusion imaging. J Am Coll Cardiol perfusion SPECT in patients without known coronary artery disease: 1999;34:777–86. incremental prognostic value and use in risk stratification. Circulation 36. He ZX, Hedrick TD, Pratt CM, et al. Severity of coronary artery 1996;93:905–14. calcification by electron beam computed tomography predicts silent 17. Hachamovitch R, Berman DS, Shaw LJ, et al. Incremental prognostic myocardial ischemia. Circulation 2000;101:244–51. value of myocardial perfusion single photon emission computed 37. Vijayaraghavan K, Singh-Khalsa M, Asher R, et al. Relationship of tomography for the prediction of cardiac death: differential stratifica- coronary ischemia by nuclear imaging and coronary artery calcification tion for risk of cardiac death and myocardial infarction. Circulation by electron beam tomography in asymptomatic patients (abstr). J Nucl 1998;97:535–43. Cardiol 2003;10:F16. 18. Hachamovitch R, Hayes SW, Friedman JD, Cohen I, Berman DS. 38. Klocke FJ, Baird MG, Lorell BH, et al. ACC/AHA/ASNC guide- Stress myocardial perfusion SPECT is clinically effective and cost- lines for the clinical use of cardiac radionuclide imaging—executive effective in risk-stratification of patients with a high likelihood of summary: a report of the American College of Cardiology/American CAD but no known CAD. J Am Coll Cardiol 2004;43:200–8. Heart Association Task Force on Practice Guidelines (ACC/AHA/ 19. Poornima IG, Miller TD, Christian TF, Hodge DO, Bailey KR, ASNC Committee to Revise the 1995 Guidelines for the Clinical Gibbons RJ. Utility of myocardial perfusion imaging in patients with Use of Cardiac Radionuclide Imaging). J Am Coll Cardiol 2003;42: low-risk treadmill scores. J Am Coll Cardiol 2004;43:194–9. 1318–33. 20. Hachamovitch R, Hayes SW, Friedman JD, Cohen I, Berman DS. 39. Rozanski A, Diamond GA, Berman D, Forrester JS, Morris D, Swan Comparison of the short-term survival benefit associated with revascular- HJ. The declining specificity of exercise radionuclide ventriculography. ization compared with medical therapy in patients with no prior coronary N Engl J Med 1983;309:518–22.